Java and backend interview questions

You isolate thread pools or connection pools per dependency so a slow search cluster cannot eat every Tomcat thread.

Naive modulo hashing remaps almost every key when N changes, causing a thundering miss on the database.

2PC holds locks across services and blocks if the coordinator dies, which kills availability.

Two requests read qty=1, both decide they can buy, both write qty=0 after insert.

Authentication filters populate SecurityContext or fail with 401 if the token is missing or invalid.

Browsers forbid wildcard origin together with Access-Control-Allow-Credentials: true.

The SQL text is parsed with placeholders

Vertical scaling buys a bigger box until GC pauses, NUMA, and single-primary database limits appear.

gRPC gives typed protobuf contracts, HTTP/2 multiplexing, and cheaper payloads for chatty internal RPCs.

Backpressure means slowing or rejecting work when downstream cannot keep up, instead of queueing until you OOM.

The tightest downstream resource wins.

The first hop generates a trace id, puts it on MDC logs, HTTP headers, and Kafka headers, and every service continues the same id.

CDC reads the database transaction log and emits row changes as Kafka events, so you cannot forget to publish after a commit.

The common product approach is a tenant_id on every table plus a mandatory WHERE and a composite primary key, enforced in code and in RLS policies.

Store a unique event id or (topic, partition, offset) in a processed table with a unique constraint, and ignore duplicates.

Cache-aside is the default: app owns the DB write and cache delete.

Add the column as nullable, deploy code that writes both old and new, backfill in batches, then add NOT NULL and drop old columns in later deploys.

Cap per phone number strictly, and cap per IP more loosely because offices and mobile NATs share addresses.

The request hits the servlet container, Spring Security filters authenticate, DispatcherServlet maps to the @RestController method, and the service runs inside a @Transactional boundary that borrows a Hikari connection.

Two nodes both believe they are primary after a partition and accept writes that cannot be merged cleanly.